Abstract:

The traditional Minimum Cross Entropy Threshold segmentation method（MCET） uses an exhaustive search form, which has the disadvantages of large computational complexity and low segmentation efficiency, which largely limits its application. Aiming at the shortcomings of the minimum cross entropy segmentation method, an improved Bat Algorithm（BA） is proposed to search the optimal solution of the threshold. Making adaptive adjustments to the weight parameters in the BA algorithm, the inertial weighting strategy that changes with the number of iterations is applied to the BA algorithm update formula, and three different improved strategies are given to solve the problem of the decline of the convergence speed of the original BA algorithm as it approaches the optimal solution. The Improved optimal BA algorithm（IBA） is applied to the minimum cross-entropy multi-threshold image segmentation. In order to explore the performance of the segmentation algorithm, it is compared with the basic BA algorithm, the Improved Particle Swarm Optimization algorithm（IPSO）, and the Fuzzy Clustering method（FC）. Experimental results show that the proposed IBA algorithm is significantly better than other algorithms in terms of operation speed and segmentation accuracy.

Key words: swarm intelligence optimization algorithm, Bat Algorithm（BA）, multi-threshold image segmentation, minimum cross entropy

摘要：

传统的最小交叉熵阈值分割法（MCET）采用穷举的搜索形式，存在计算复杂度大、分割效率低的缺点，在很大程度上限制了该方法的应用。针对最小交叉熵分割法存在的不足，提出采用改进蝙蝠算法（BA）来搜索阈值的最优解。对BA算法中的权重参数做自适应调整，将随着迭代次数变化而变化的时变惯性权重策略应用于BA算法更新公式，给出三种不同的改进策略解决原始BA算法在靠近最优解时收敛速度下降的问题。将改进后的最优BA算法（IBA）应用于最小交叉熵多阈值图像分割中，与基本BA算法、改进的粒子群优化算法（IPSO）、模糊聚类方法（FC）三种方法进行对比性实验。实验结果表明，提出的IBA算法运算速度和分割精度效果明显提升。

关键词: 群智能优化算法, 蝙蝠算法（BA）, 图像分割, 最小交叉熵